On a sunny, breezy day in the WindRiver Mountains of Wyoming, agroup of geologists are peeringintently at a dark ridge of rock. Some

2.7 billion years ago, these rocks were alive withvolcanic fire. Today, they jut out of a mountain-side like the spiny tail of a sleeping dragon.

This rock, says Kevin Chamberlain, a geol-ogist from the University of Wyoming inLaramie, could be a special kind — a lavacalled a komatiite. Today, Earth’s interior is toocool to produce this particular rock; 2.7 billionyears ago, the hot lava would have run likewater over the barren landscapes.

But as the other geologists chip off fresh lay-ers and scrutinize them through hand lenses,murmurs of dissent start to grow. Most geolo-gists have never seen a komatiite; they arefound almost exclusively among rocks of theArchaean era, which are more than 2.5 billionyears old and thus very rare. But these menand women are experts in the truly old. Andon the Wyoming hillside few of them are con-vinced that they are seeing the rock texturestypically found in komatiites.

The scene brings home the difficulties of try-ing to study the early Earth — there aren’t manyold rocks to look at, and those that are aroundare often so altered, chemically and physically,as to be nearly indecipherable. But as if to rec-ompense those who study them, such ancientrocks, particularly of Archaean age, offer geolo-gists great rewards. It is in the Archaean that thefirst earthly ecosystems are found, with theirclues to life’s earliest days on the planet. And it isin the Archaean that scientists can look for thebeginnings of plate tectonics.

Plate tectonics is the grand unified theory ofgeology. Everything we see today, from theabyssal plains of the oceans to the heights ofthe Himalayas, is shaped by plate tectonics. Asfar back as there has been complex life — andperhaps even before — continents have cometogether and moved apart in a dance that hasaltered climates and geographies, opening upnew possibilities for life and sometimes closingdown old ones.

But it may not always have been so. Plate tec-tonics is driven by Earth’s heat and constrainedby the physical and chemical properties of thecrust and mantle. The further back in time yougo, the more different these things are likely tohave been. It’s been argued that on the earlyEarth, crustal plates, floating on a heat-soft-ened layer of material beneath, would havesimply been too thick and buoyant to getdragged beneath each other as they are today.And the greater temperature of the early Earth’sinnards would probably have made them move

in very different patterns from those typical oftoday’s tectonics.

On other Earth-like planets there’s no evi-dence for today’s plate tectonics. Planets donot have to work this way, and there was prob-ably a time when this one didn’t. “You don’tjust make a silicate planet and plate tectonicsstarts,” says Robert Stern, a geologist at theUniversity of Texas, Dallas. “Something spe-cial has to happen.”

Dynamic planetThe nature of that special something cuts tothe discipline’s philosophical heart. Since theearly nineteenth century, geology has beenruled by the principle of uniformitarianism —that the planet operates on unchanging laws,and that the present can be used as a key to thepast. But how can that approach hold up whena science from a world where plate tectonics

explains more or less everything is applied to aworld that may have lacked it? How can youunderstand ancient rocks when you do notknow what processes put them there?

The geologists clustered around the possi-ble komatiites in the Wyoming hills had gath-ered to discuss these questions. Their visit tothe mountain ridge had been organized aspart of a June conference held in Lander,Wyoming. The specific aim of the meetingwas to try to fix a date for the onset of platetectonics: the earliest possibility is prettymuch straight after the planet formed, about4.5 billion years ago; the latest is just 1 billion years ago.

To help them decide, the scientists broughtto the table data from an array of disciplines.Geochemistry can help clarify the temperatureand pressure at which Archaean rocks formed.Fragments of zircon crystals dated even earlier

NEWS FEATURE NATURE|Vol 442|13 July 2006

128

Plate tectonics has created oceans and pushed up mountain ranges. But when did the process that shapes the planet get going? Alexandra Witze joins the geologists debating the issue.

The start of the world as we know it

Ancient rocks, such as those in the Wind River Mountains, could help determine when plate tectonics began.

A. W

ITZ

E

13.7 plate tectonics MH 11/7/06 10:29 AM Page 128

Nature Publishing Group ©2006

NATURE|Vol 442|13 July 2006 NEWS FEATURE

129

— to the Hadean eon, which stretches fromabout 3.8 billion years ago to the planet’s birth— can provide hints about what Earth’s surfaceenvironment was like back then. Palaeomag-netic studies can show how land massesmoved across latitudes. And structural geol-ogy can identify features that, in today’s worldat least, seem to be indicative of plate tectonics.But in all these approaches, as with the komati-ites, age makes the picture hard to discern.

Crashing continentsScant and difficult-to-interpret evidence pre-sents one set of problems; slippery definitionspresent another. Plate tectonics has lots of con-stituent parts. It’s not just a theory of howthings move, but of how they are made andfrom what. For example, explanations for dif-ferent sorts of volcanism in different settingsalso explain why the mineral make-up of con-tinental crust and the crust beneath the oceansis so different.

Working out which attributes are essentialto the theory, and which incidental, is not easy.The 65 attendees at the Wyoming conferencecame up with 18 different definitions of platetectonics. Three components, most agreed,were key: there must be rigid plates at the sur-face of the Earth; those plates must move apartthrough ocean spreading, with new crustbeing made where the sea floor pulls apart;and the plates must on occasion dive beneatheach other at subduction zones (see graphic).

The problem is that Earth could display oneor even two of these properties without neces-sarily having a system like that described bymodern plate tectonics (see ‘A world withouttectonics’). Take rigid plates. Palaeomagneticand other studies show that sections of Earth’scrust moved relative to each other in theArchaean, just as modern crustal plates do.But ice floes on a polar sea move in the sameway, points out geophysicist Don Anderson ofthe California Institute of Technology inPasadena — and those ice floes aren’t experi-encing plate tectonics.

Of the three, it seems subduction is closestto being diagnostic of plate tectonics. Subduc-tion is the process by which one crustal plate

slips beneath another, to be recycled into themantle. Subduction requires rigid plates, andas it involves the destruction of crust, newcrust must be created elsewhere, presumablyat oceanic spreading ridges (see graphic); oth-erwise, continental crust would eventually dis-appear. Some argue that this means platetectonics should date further back than theearliest firm evidence for subduction.

A dramatic use of this argument is thatmade by Stern. In a paper published last year,he took an extreme position, proposing thatEarth has been free of plate tectonics foralmost four-fifths of its life, with the system we

see today starting up only a billion years ago1.He had two lines of argument. One was thatplate tectonics could not begin until Earth’scrust was cool enough, and that barrier wasonly passed about a billion years ago. Theother was that the only reliable evidence forsubduction on the early planet came from aperiod more recent than that.

Stern points to the geological record of threetypes of rock. Ophiolites are distinctive sec-tions of the ocean crust that gets mashed up,often through subduction, on the edges ofcontinents. Stern argues that very few of theserocks are more than a billion years old. Meta-morphic rocks called blueschists, produced bysqueezing the basalt from which oceanic crustis made at high pressures but not very hightemperatures, are being made in today’s sub-duction zones; none, Stern says, has beenfound that is more than 800 million years old.And rocks from ‘ultra-high-pressure terranes’of the sort produced where one plate rides overanother are at most 630 million years old.

He also makes a more general point. A dra-matic shift, such as the introduction of platetectonics, must have had huge planetary conse-quences. And between 780 million and 580

million years ago, Stern says, there was a seriesof glaciations, some very extreme — giving riseto the term ‘snowball Earth’. “It was a wild timeof change,” says Stern. “The biosphere was outof control.” On the basis that dramatic effectsrequire dramatic causes, he argues that theintroduction of plate tectonics, and with it anincrease in planet-cooling volcanic eruptions,might have precipitated the great glaciations.

An age gone byAfter reading Stern’s arguments, AlfredKröner of the University of Mainz in Germanyfired off a rebuttal. He argues that there’splenty of evidence for plate tectonics stretch-ing back at least 3.1 billion years2 — includinggeochemical work, seismic images of the‘sutures’ where colliding continents join and,indeed, a few ancient ophiolites. “I believe wecan see these features all the way back” — pos-sibly all through the Archaean, says Kröner.

The exchange of papers led to the Wyomingconference. “It was overdue,” says Kröner.“Nobody ever talks to one another.” InWyoming, they did: palaeomagnetists clus-tered around a white board with field geolo-gists; geophysicists sat down for a beer with geochemists.

Some of the newly shared data favour anearly start for plate tectonics. Geoff Davies, amodeller at Australian National University inCanberra, presented work suggesting that oneof the biggest stumbling blocks to an early startmay have been removed. In the early 1990s,computer models created by Davies and oth-ers suggested that the crust on the early Earthwould have been too thick and buoyant to getdragged down beneath another plate duringsubduction. But new simulations, using moresophisticated calculations, suggest that thecrust may have been thinner than oncethought3 — as thin as 4 kilometres or less —which would be thinner than today’s crust.“Maybe plate tectonics on the early Earth wasviable after all,” says Davies.

In other cases, recent findings overturnedevidence for early plate tectonics. In 2001, ateam reported that an ophiolite from Dong-wanzi, China, was 2.5 billion years old — mak-

Subducting oceanic plate Passive continentalmargin (no subduction)

Mantleupwelling

Oceanicspreading ridge

Continentalplate

Volcano

THE DRIVING FORCES OF PLATE TECTONICS

“You don’t just make a silicateplanet and plate tectonics starts.

Something special has to happen.”— Robert Stern

13.7 plate tectonics MH 11/7/06 10:29 AM Page 129

Nature Publishing Group ©2006

NEWS FEATURE NATURE|Vol 442|13 July 2006

130

ing it by far the oldest such subduction rem-nant yet discovered4. Now Guochun Zhao, ofthe University of Hong Kong, has re-datedthose rocks, giving them an age of just 300 mil-lion years.

Timothy Kusky of St Louis University inMissouri, who led the original study, says thatZhao took samples from a part of the rockalready known to be much younger than themain part of the ophiolite. But several atten-dees at the meeting said they found Zhao’s dataconvincing. If true, it would pull the earliestevidence for ophiolites at least half a billionyears towards the present, leaving theArchaean an ophiolite-free zone.

The Chinese ophiolite isn’t the only evi-

dence that is getting fresh scrutiny. For a whiletwo independent groups have been quietlywarring over the significance of a pile ofancient zircons from the Jack Hills region ofWestern Australia. The zircons are crystalsthat formed in the Hadean and later becameincorporated into younger rocks.

Last year in Science5, geochemist Mark Har-rison of the University of California, Los Ange-les, and colleagues used the Jack Hills zircons toargue that continental crust was present 4.4 bil-lion to 4.5 billion years ago. The evidencecomes in the form of hafnium isotope ratios inthe zircon crystals, which preserve signals ofthe lighter minerals typical of continental crust.The data also suggest, Harrison argues, that

that crust was being recycled down into themantle by 4.4 billion years ago — perhapsthough a process similar to plate tectonics.

Simon Wilde of the Curtin University ofTechnology in Perth, Australia, isn’t so sure.“You have to be very careful with these rocks,”he says. Measuring one spot on a crystal, asopposed to another, can yield very differenthafnium values that lead to very different inter-pretations, he says. Wilde argues that the zir-cons should be interpreted more conservatively— that the evidence points to there being somecontinental crust, but not plate tectonics and itsassociated recycling, by 4.4 billion years ago6.

Ground forcesSuch differences of interpretation make theproblem of solving when plate tectonics beganextremely difficult. In many cases, data can beinterpreted in several completely differentways — all of which may seem valid.

For instance, another Australian geologistpresented seemingly convincing evidence thatplate tectonics had begun by 3.3 billion yearsago in Western Australia, based on the verydifferent histories of two sections of an ancientrock formation called the Pilbara. HughSmithies of the Geological Survey of WesternAustralia says that the eastern part of the Pil-bara, between 3.5 billion and 3.2 billion yearsold, “shows no clear evidence for modern-styleplate tectonics”. It contains some geochemicalmarkers that suggest subduction, but theycould just as easily be explained by hotupwellings of rock known as mantle plumes orother non-tectonic phenomena.

In contrast, looking at the western part of thePilbara — which is 3.3 billion to 3.0 billion yearsold — Smithies sees plenty of evidence for platetectonics. There are geochemical signatures thatcannot be explained by other factors, and the

Clues to the past: zircon crystals (inset) in the Jack Hills of Western Australia have been used to argue for an early start to plate tectonics.

Time to talk: Earth scientists gathered at a meeting in Wyoming to present diverse data on the early Earth.

M. H

ARR

ISO

N

A. J

. VA

N D

ER V

ELD

EN

13.7 plate tectonics MH 11/7/06 10:29 AM Page 130

Nature Publishing Group ©2006

131

rocks show features that hint that plates hadinteracted along their edges. Smithies thinks thewestern Pilbara contains the remains of anoceanic arc — the sort of line of islands, such asthe Aleutians of Alaska, that are characteristicof some oceanic subduction zones7.

But then along came Julian Pearce of CardiffUniversity in Wales, who argued that each ofthe geochemical markers in the western Pil-bara can be explained by other phenomena,

such as magmas with an unusual amount ofwater in them, or crustal material from differ-ent places getting mixed up. The variousresearchers are hoping to settle the matter witha field trip. An excursion is already planned fornext year, to re-examine the evidence for platetectonics in the western Pilbara.

Field trips don’t always resolve things. In theWind River Mountains, the meeting attendeescontinued to argue about plate tectonics asthey hiked from outcrop to outcrop. But aweek of communing at the conference andunder the high mountain sun brought themtoward a consensus of sorts.

Meeting organizers polled the attendeestwice on when they thought plate tectonicsbegan. At the beginning of the meeting, guesseswere spread over more than 3 billion years ofEarth history. At the end, a closing ballotshowed that many had begun to push theirthinking further back into the past; a majorityof attendees voted for plate tectonics havingstarted between 3 billion and 4 billion years ago.

Kent Condie, one of the meeting organizers,calls that a success. “We’ve got a majorityfavouring a definition and approach,” saysCondie, of the New Mexico Institute of Miningand Technology in Socorro, New Mexico.“Sure, there will be a minority point of view.”

At the conference, that minority prettymuch constituted Stern. By the end of themeeting, he remained the one person votingfor a start to plate tectonics at 1 billion yearsago. “It’s not a simple question,” he maintains.And on that, at least, others agree.

Michael Brown, a geologist at the universityof Maryland in College Park, doesn’t endorseStern’s late start. But he does think that thenature of plate tectonics changed around thattime. In a paper in press in Geology8, Brownsuggests that there have been two styles of

plate tectonics: the modern kind that we seetoday, and an earlier version that lasted fromabout 2.7 billion to 700 million years ago. Evi-dence for the earlier style, he says, comes fromminerals that are typical of higher-tempera-ture, lower-pressure environments; these sug-gest a hotter Earth where plates did notsubduct beneath each other to great depthsand pressures. Minerals characteristic of high-pressure environments typify the later style.The properties of these minerals suggest tohim that true plate tectonics, in which oneplate subducts deeply beneath another, did notbegin until about 700 million years ago.

And there is a possible further complica-tion. Geophysicist Paul Silver, of the CarnegieInstitution of Washington, raised the notionthat plate tectonics may have started andstopped several times during Earth’s history.This is also an idea that Stern is comfortablewith — he uses it to explain the presence of asmall number of ophiolites about 2 billionyears ago.

An ‘intermittent approach’ would be a won-derful way to reconcile things — but it takesgeology even further from the comfortingrealm of uniformitarianism, into a worldwhere the most basic principles come and goin fits and starts. ■

Alexandra Witze is Nature’s chief ofcorrespondents for America.

1. Stern, R. J. Geology 33, 557–560 (2005).2. Cawood, P. A., Kröner A., Pisarevsky, S. GSA Today 16, 4–11

(2006).3. Davies, G. F. Earth Planet. Sci. Lett. 243, 376–382 (2006).4. Kusky, T. M., Li, J-H. & Tucker, R. D. Science 292, 1142–1145

(2001).5. Harrison, T. M. et al. Science 310, 1947–1950 (2005).6. Valley, J. W. et al Science 312, 1139a (2006).7. Smithies, R. H. et al. Gondwana Res. (in the press).8. Brown, M. Geology (in the press).9. Bédard, J. Geochim. Cosmochim. Acta 70, 1188–1214

(2006).

Slow work: geologists hunt for structural signatures in rocks that can only be explained by plate tectonics, to try to identify when the process started.

NATURE|Vol 442|13 July 2006 NEWS FEATURE

A. P

OLA

T

With so much uncertainty over Earth’s earlyhistory, not many geologists are willing toimagine entirely different ways in which theworld might have worked before platetectonics. One of the brave few is Jean Bédard, a geologist at the Geological Survey of Canada.

At a June conference held in Lander,Wyoming, Bédard presented one of the fewfully worked out accounts of how a pre-plate-tectonic Earth might have worked. In hismodel, hot upwellings of rock known asmantle plumes partly melt the crust abovethem. This melting distills the crust, producingmaterial from which light, continental-stylecrust is made.

The material left behind as the melt iscreamed off — denser than it was before thedistillation — then detaches itself from thecrust and sinks back into the mantle. There, itwould mix with more mantle material — andthe whole process would start all over again9.

Bédard himself admits that he has no idea ifhis proposal is right. But it is, he says, adetailed alternative theory to plate tectonics, and one that can be tested with further studies. A.W.

A WORLD WITHOUTTECTONICS

13.7 plate tectonics MH 11/7/06 10:29 AM Page 131

Nature Publishing Group ©2006